← → clavier · espace · tactile
Confidential — Executive Briefing
A
ALIA Santé

Accelerating Lilly's Clinical Pipeline
with Hybrid AI Trials

Virtual Patients · Hybrid Protocols · FDA-Ready Validation

−40%
Patient enrollment
−18 mo
Trial timeline
+15%
Trial success rate
Context

Clinical Development Is Broken

The traditional model wastes resources that Lilly can reallocate to pipeline acceleration.

$2.6B
Average cost per approved drug
Phase 3 alone consumes 40–60% of total development budget, mostly in patient recruitment.
80%
Trials delayed by enrollment
Patient scarcity in rare oncology targets (SMARCA2, PTK7) and complex metabolic indications (MASH, HFpEF) is the #1 blocker.
12–15 yrs
Average time to market
Every 6 months of delay = ~$500M in lost peak sales for a blockbuster candidate.
30%
Phase 3 trials fail to enroll on time
Running each new trial as if it's the first time we've studied that indication is inefficient — we know so much about controls already.
💡
"Would you be willing to die to give a p-value?" — Janet Woodcock, FDA, 2019. ALIA's answer: no. Use virtual patients.
Technology Deep-Dive

ALIA's Generative Neural Architecture

A Neural Boltzmann Machine trained on harmonised multi-source clinical data to generate probabilistic patient trajectories.

Input Data
📋 Past RCT control arms
🏥 Observational studies
🧬 Registries & RWE
Multi-source · Multimodal · Harmonised
→
ALIA Neural Engine
⚙️ Neural Boltzmann Machines — learn full joint probability distribution P(outcomes, time)
🔀 Generative model: baseline snapshot → probabilistic future trajectory
📊 Handles sparse, heterogeneous, multimodal clinical data (labs, biomarkers, scores, images)
Trained · Validated · Per indication
→
Output
👤 Virtual Patient
📈 Full trajectory prediction (50+ variables)
✅ Statistically indistinguishable from real data
High-fidelity · Dynamic · Updatable
🧠
Learns from history
Trained on 10,000+ subjects per indication from prior trials — every prediction grounded in real disease biology.
🎯
Patient-specific
Conditioning on baseline yields an individual digital twin — not a population average but a personalised forecast.
🔄
Dynamically updatable
As new real data arrives during the trial, the virtual twin is refreshed — staying calibrated to actual trajectory.
Methodology

How a Hybrid Clinical Trial Works

PROCOVA (Prognostic Covariate Adjustment) — EMA-qualified, FDA-accepted — uses virtual patients to boost statistical power.

1
Historical Data Harmonisation
ALIA curates multi-source datasets (past RCTs, observational, registries) for the target indication — aligning variables, visit cadences, and inclusion criteria.
2
Digital Twin Generator (DTG) Training
ALIA trains a Neural Boltzmann Machine per indication. Model is validated on held-out data — correlation between predicted and observed outcomes is the key quality metric.
3
Trial Design — Hybrid Cohort
For each enrolled patient, ALIA generates their virtual twin at baseline. The trial runs normally — treatment vs. placebo — but with a smaller control arm, because digital twins provide additional statistical signal.
4
PROCOVA Analysis — Virtual Twins as Super-Covariates
Instead of adjusting for age/sex, the analysis adjusts for each patient's predicted outcome (their digital twin). This "super-covariate" captures variance from prior data, boosting effective sample size without adding real patients.
5
Result: Same Power, Fewer Patients
Example (AD trial): Standard 1,600 pts → Hybrid 1,433 pts with same 80% power, saving $59M and 2 months enrollment — or alternatively boosting power to 84.6% with same sample size.
Regulatory

FDA & EMA Validation Pathway

ALIA's Quality Report de-risks the regulatory conversation before the first patient is enrolled.

✅ EMA Qualification
PROCOVA officially qualified by EMA
EMA issued formal qualification opinion for PROCOVA as a valid statistical method for Phase 2 & 3 RCTs. Digital twins recognised as legitimate prognostic covariates.
✅ FDA Concurrence
FDA concurred with EMA's PROCOVA qualification
FDA draft guidance "Considerations for the Use of AI to Support Regulatory Decision-Making" (2024) explicitly covers AI-generated digital twins in clinical submissions.
📄 ALIA's Quality Report — The Regulatory Dossier We Build With Lilly
Model card
Architecture, training data, validation metrics per indication
Context of use
Explicit COU statement linking virtual patients to protocol endpoints
Traceability & audit trail
Full data lineage — from source dataset to generated virtual patient
→ ALIA accompanies Lilly through the pre-submission meeting with FDA — the Quality Report is the submission-ready package for the hybrid protocol.
Our Offer

What ALIA Delivers to Lilly

Two tightly integrated deliverables — science and regulatory packaged together.

👤
Deliverable 1 — High-Fidelity Virtual Patients
Synthetic patient cohorts generated by ALIA's Neural Boltzmann Machines, calibrated to the specific indication, protocol, and inclusion/exclusion criteria of each Lilly study.
✦ Designed for specific study COU — not off-the-shelf
✦ 50+ predicted variables per patient (labs, scores, biomarkers)
✦ Probabilistic — captures individual uncertainty
✦ Updatable during trial as new data arrives
✦ Augments — never replaces — real-world evidence
📑
Deliverable 2 — Hybrid Protocol Validation Dossier
ALIA co-authors the full regulatory package for the hybrid clinical study design — ready for FDA pre-submission meeting and IND/NDA inclusion.
✦ SAP (Statistical Analysis Plan) with PROCOVA methodology
✦ Model card + traceability + audit trail
✦ Context of Use statement per endpoint
✦ Sample size justification with power simulations
✦ Support through FDA pre-submission and Q&A
🤝
Engagement model: ALIA embeds alongside Lilly's clinical team from protocol design through analysis — not a vendor, a scientific partner with regulatory skin in the game.
Priority Programs

Lilly Oncology — ALIA's Highest-Value Targets

Rare targets and ADCs where ALIA's hybrid approach converts me-too into first-in-class status.

Program
ALIA Time Gain
First Mover
Market Potential
LY4175408 — ADC PTK7
Rare target · race open vs AZ/ImmunoGen
+6–12 months
🔴 Very High
$1.2–2.4B
LY4050784 — SMARCA2 inhibitor
Ultra-rare target · no competitor with AMM
+6–12 months
🔴 Very High
$600M–1.5B
LY3962673 — KRAS G12D (niches)
Emerging market · small denominator → ideal for hybrid
+6–9 months
🔴 Very High
$800M–1.3B
Sofetabart — ADC FR post-Elahere
Best-in-class positioning window
+6–9 months
🟡 High
$700M–1.4B
Olomorasib — CRC 2nd line
Sub-indication hybrid arm
+4–6 months
🟡 High
$250–450M
→ Combined oncology first-mover potential: $4–8B vs. me-too status without acceleration.
Franchise Growth

Lilly Metabolism & Neurology — Mega-Franchise Value

Where 6–12 months of acceleration translates into tens of billions in cumulative additional revenues.

🧬 Metabolism
SYNERGY-NASH — Tirzepatide MASH 🔴 Very High
1,500–2,000 pts, liver biopsy, strict fibrosis criteria → hybrid −30–40%, −9–12 months. Market: $4–8B if best-in-class.
TRIUMPH — Retatrutide Obesity 🔴 Very High
Triple agonist, tight race vs Novo/BI. −9–12 months seals first-in-class. Market: $15–25B.
Orforglipron ATTAIN — Oral GLP-1 Obesity
First oral GLP-1 obesity. Hybrid + RWE secures superiority vs Rybelsus. Market: $12–20B.
💰 ROI: $50–80B additional revenues 2027–2035 · $300–500M cost savings per Phase 3
🧠 Neurology & Immunology
Donanemab TRAILBLAZER-ALZ 3 Prevention 🔴 Very High
Asymptomatic, 5-yr follow-up, slow enrollment. Hybrid + RWE registries: −9–12 months. Market: $6–12B.
Omvoh Combinations (eltrekibart / MORF-057) 🔴 Very High
First IL-23 + anti-neutrophile or oral α4β7 combo. −9–12 months. Market: $9–18B cumulative.
Olumiant Pediatric Alopecia
First JAK approved pediatric AA before Pfizer 2026–27. Data augmentation pediatric populations. Market: $1.5–3B.
💰 ROI: $30–50B additional revenues 2027–2035 · $200–400M cost savings
Strategic Value

The Competitive Advantage ALIA Generates for Lilly

From "me-too" to "first-in-class" on 5+ programmes — a structural portfolio advantage worth $80–130B.

5+
Programmes converted to first-in-class
6–12
Months gained per priority programme
−40%
Patient enrollment per hybrid trial
$500M–1B
Cost savings Phase 3 portfolio
🏁 First-Mover Economics
In rare oncology (PTK7, SMARCA2, KRAS G12D niches), the first NDA filed wins the lion's share. At $500M/year in lost revenues per 6-month delay, ALIA pays for itself in weeks of acceleration.
📊 Regulatory De-risking
ALIA's Quality Report removes the single biggest uncertainty in hybrid trial adoption — FDA/EMA acceptance. Lilly enters pre-submission meetings with a validated methodology, not a request for guidance.
🧬 Enrollment in Rare Populations
For SMARCA2, KRAS G12D, pediatric AA — patient scarcity is the limiting factor. ALIA's hybrid cohorts remove the bottleneck, making trials feasible that would otherwise require 3–5 years of enrollment.
💊 Metabolism Franchise Defence
On Tirzepatide MASH and Retatrutide, Lilly leads — ALIA extends that lead vs Novo/BI by 9–12 months per programme, compounding into franchise dominance worth $50–80B through 2035.
Proposed Engagement

Start Small. Win Fast. Scale Together.

A 3-phase partnership roadmap — from proof of concept to portfolio-wide deployment.

Phase 1 — Q2–Q3 2026
Proof of Concept
Select 1–2 priority programmes (recommended: LY4175408 ADC PTK7 + SYNERGY-NASH). ALIA delivers virtual patient cohort + preliminary Quality Report within 90 days.
Phase 2 — Q3–Q4 2026
FDA Pre-Submission
Co-author hybrid protocol dossier with Lilly regulatory team. Joint FDA pre-submission meeting preparation. Finalise SAP and power simulations. Enrolment starts with hybrid design.
Phase 3 — 2027+
Portfolio Deployment
Scale across 5–8 Lilly priority studies across Oncology, Metabolism, Neurology. Embedded ALIA team within Lilly CDO. Dedicated indication-specific model development.
90 days
To first deliverable
$0 upfront
PoC milestone-based
1 sponsor
ALIA from protocol to NDA
1 / 10